AMD cuts ray tracing memory for foliage by 98 percent

4 min read
AMD cuts ray tracing memory for foliage by 98 percent

Why foliage taxes ray tracing memory

Modern games rely on ray tracing to simulate realistic light interaction. While reflections on metal and glass are obvious, the countless leaves, branches and grass blades create a hidden memory burden. Each tiny surface must store ray data, acceleration structures and shading information. In large open worlds this adds up quickly, forcing developers to lower settings or sacrifice detail.

Traditional approach and its limits

Historically, developers have used simplified collision meshes or baked lighting for foliage. When full ray tracing is enabled, the GPU must allocate separate buffers for every instance. This leads to memory spikes that can exceed the capacity of even high‑end cards, causing frame drops or stutters.

AMD’s breakthrough in memory efficiency

At the recent Radeon developer summit, AMD unveiled a set of hardware and driver enhancements that reduce the memory required for ray traced foliage by roughly ninety eight percent. The improvement stems from three key innovations:

  • Instance compression. AMD’s new engine packs similar foliage instances into shared data blocks, eliminating redundant storage.
  • Dynamic level of detail for acceleration structures. The GPU builds coarse hierarchies for distant trees, refining them only when the player approaches.
  • On‑chip caching. A dedicated cache stores frequently accessed ray data, cutting the need to read from main VRAM.

These techniques are implemented in the RDNA3 architecture and exposed through the latest driver release. According to AMD’s technical brief, the memory footprint for a typical forest scene drops from several gigabytes to under a hundred megabytes.

Real‑world impact on game titles

Early testing with titles that heavily feature vegetation, such as Horizon Forbidden West and Microsoft Flight Simulator, shows noticeable gains. Frame rates improve by five to ten percent on a Radeon RX 7900 XTX, while power draw falls by roughly three watts during dense foliage sections.

Developers also report that the lower memory demand frees up space for higher resolution textures or additional ray traced effects, creating a net win for visual fidelity.

How the technology works under the hood

At the core of AMD’s solution is a redesign of the Bounding Volume Hierarchy (BVH) that underpins ray tracing. Instead of generating a separate BVH for each tree, the GPU now creates a unified BVH that references shared geometry. This reduces the number of nodes that must be stored.

The driver also introduces a dynamic BVH update algorithm that adjusts hierarchy depth based on camera distance. Near objects retain full detail, while far objects use a simplified structure that consumes far less memory.

Comparison with competing solutions

While NVIDIA has long promoted its RTX hardware for ray tracing, AMD’s memory reduction targets a different pain point. NVIDIA’s recent memory compression features focus on texture data rather than acceleration structures. AMD’s approach complements those efforts by tackling the BVH itself.

Both companies aim to make ray tracing viable at higher resolutions, but AMD’s method may prove more scalable for open world games where foliage dominates the scene.

Implications for developers and gamers

For developers, the reduced memory requirement means fewer compromises when implementing full ray tracing. Studios can allocate VRAM to other assets, such as higher resolution skyboxes or more detailed character models. The new tools integrated into the Radeon GPU Profiler also provide visibility into memory usage, helping teams fine‑tune performance.

Gamers benefit from smoother gameplay without needing to manually lower foliage settings. On mid‑range systems, the lower VRAM demand can prevent out‑of‑memory crashes that previously forced players to disable ray tracing altogether.

Future outlook

AMD indicates that the technology will be extended to other geometry‑heavy assets, including crowds and destructible environments. If the same memory efficiencies can be achieved for those elements, the overall burden of ray tracing could shrink dramatically.

Industry analysts from the University of California research on real time ray tracing suggest that such hardware‑level optimizations are essential for the next generation of immersive games.

In the meantime, players can expect upcoming patches for existing titles to incorporate AMD’s driver updates, delivering immediate performance gains on compatible hardware.

Comments

No comments yet. Be first.

More from this author